New Approach To Identify Analogous Reservoirs

Abstract

Identifying analogous reservoirs is important in planning the development of a new field. Usually, information available about a new area is limited or even nonexistent. Traditionally, the search for analogous reservoirs is carried out by experienced geoscientists. This search is subject to the availability of expertise, and the results heavily depend on the geology of the area. This paper presents a systematic and unbiased procedure to search for analogous reservoirs on the basis of information contained in a large validated database of engineering and geologic parameters. Each reservoir has its own "fingerprint" characterized by a set of properties, which commonly vary from one reservoir to another. The method uses multivariate statistical techniques to find a unique and reproducible list of reservoirs with fingerprints that are most similar to the selected target. The flexibility of the method allows for evaluation of different scenarios [e.g., static, dynamic, pressure/volume/temperature (PVT) behavior] by analog class.This method consists of four steps: data preprocessing, keyparameters selection, multivariate analysis, and similarity ranking. The first step involves analysis and preprocessing of the data.With key-parameter Selection, variables with largest impact on the case to be evaluated are identified. The third step, multivariate analysis, applies several multivariate techniques such as principal-component analysis (PCA) and cluster analysis. Finally, in the similarity-ranking step, we apply a similarity function to the group of previously selected "analogous reservoirs," generating a similarity ranking of analogous reservoirs.Casablanca oil field was used as a target reservoir to validate this new method. This reservoir is a mature carbonate field very well-known by Repsol, which had experts that identified four analogs. The new developed method was independently applied in this case to obtain 19 analogous reservoirs sorted by similarity criteria. The maximal similarity found was 85% for the Amposta Marino reservoir; this was one analogous reservoir independently identified by the business-unit team. Moreover, these four analogous reservoirs previously identified were within the first ten positions in similarity ranking. These results are encouraging because they ensure a reproducible response regardless of the user expertise.The singular feature of this new method is that it is based on a similarity function, which accounts for all the weighted key parameters (KPs) simultaneously. Commercial software often uses sequential filters. As a result, the procedure we present will support the predictive search of missing properties for the target reservoir, reducing the uncertainty for decision making.

Hybrid and Automated Machine Learning Approaches for Oil Fields Development: the Case Study of Volve Field, North Sea

“…In this way, we can prevent data leaks during the training process and we are able to correctly analyze the predicted oil reservoir. For cases with more diverse datasets that contain information about several reservoirs, the analogs matching procedure [33] can be applied to use different reservoirs for training and validation of the model.…”

A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs

“…Thus, subsurface resource analog selection, the choice of what analog to use across various scales and study intervals, is important to constrain subsurface resource model input univariate, multivariate, and spatial distributions. Although there are numerous uses of analogs in the subsurface, a couple prominent ones are the use of production data from existing wells as analogs for future wells to generate a type-curve for well hydrocarbon production over time during prospect evaluation and field estimation (Sidle and Lee, 2010; Rodríguez et al, 2014; Sun and Pollitt, 2022), and mining mineral deposit analogs help identify new prospecting targets (Prior et al, 2021). Accurate subsurface resource analog selection is crucial because an inappropriate analog can lead to erroneous information to populate models and become a source of imprecise and inaccurate subsurface uncertainty model predictions.…”

“…However, this distance-based approach is disrupted by the previously discussed distance distortion in a higher dimensional space. Rodríguez et al (2014) address distance distortion in high-dimensional space with principal component analysis (PCA) for dimensionality reduction along with avoidance of multicollinearity between reservoir features (Jollife, 2002; Rodríguez et al, 2014; Jollife and Cadima, 2016). Olukoga and Feng (2022) apply heuristic algorithms, including k-means, k-medians, and hierarchical clustering algorithms, to identify analogous miscible CO 2 flooding projects after applying PCA for dimensionality reduction coupled with a similarity measure using weighted Euclidean distance.…”